What is Milvus?
Vector database for RAG and semantic search that runs on a laptop, a server or a Kubernetes cluster.
Who is Milvus best for?
Milvus is best for AI engineers who need vector search over billions of embeddings on their own servers.
RAG and semantic search
Milvus stores embeddings next to integers, strings and JSON fields, so a search can filter on metadata in the same query.
Keyword plus vector search
BM25 full text search and sparse embeddings such as SPLADE sit beside dense vectors, so hybrid search needs no second engine.
Prototypes that grow
Milvus Lite runs in a notebook after a pip install, and the same client code later points at a Standalone server or a cluster.
Very large collections
Distributed mode splits compute from storage on Kubernetes and scales out to tens of billions of vectors.
How much does Milvus cost?
Milvus is free to self-host under Apache 2.0; Zilliz Cloud, the managed version, has a free tier with 5 GB of storage, pay-as-you-go serverless clusters and dedicated clusters from $126 a month.
- Apache 2.0, no limits on vectors or collections
- Lite, Standalone or Distributed deployments
- You run upgrades, backups and scaling
- 5 GB of storage
- 2.5M vCUs a month
- Up to 5 collections, community support
- Serverless clusters from $0, billed by use
- Backup, restore and basic monitoring
- 30-day free trial
- 99.95% uptime SLA
- SSO, audit logs and granular RBAC
- Private endpoints and VPC peering
Zilliz Cloud is the fully managed Milvus from Zilliz, the project's main contributor. Business Critical (HIPAA-eligible, customer-managed keys, global clusters) and bring-your-own-cloud deployments are priced on request. Dedicated cluster capacity starts at $63 per million vectors a month on performance-optimized nodes, $16 on capacity-optimized and $5 on tiered storage, for 768-dimension vectors. Full pricing on zilliz.com. We checked these prices on .
How do readers rate Milvus?
No one has rated Milvus yet. Be the first.
Sign in to rate. One rating per account.
Is Milvus open source?
Yes. Milvus's code is public under the Apache 2.0 licence, so you can run it on your own server.
On GitHub
46,649stars
+272 (+0.6%) in the last 30 days
- Last commit
- 12 Nov
- Repository age
- 7 years
- Latest release
- 11 Nov
- Licence
- Apache 2.0
- Self-hosted
- Yes
- Repository
- milvus-io/milvus
What are the best Milvus alternatives?
Milvus is an open source alternative to Pinecone. Qdrant, Supavec, Deeplake and HelixDB do the same job and are open source too.
Managed vector database for search, RAG and agents.
Open-source vector search with real-time indexing.
Open-source RAG for your own data.
Database for AI data that keeps vectors with images, video and text, plus search and versioning.
Graph database with built-in vector and full-text search, made for AI agent memory and retrieval.
Questions about Milvus
What is Milvus?
Vector database for RAG and semantic search that runs on a laptop, a server or a Kubernetes cluster.
Who is Milvus for?
Milvus is best for AI engineers who need vector search over billions of embeddings on their own servers.
Is Milvus free?
Yes. Milvus has a free Self-hosted plan: apache 2.0, no limits on vectors or collections.
How much does Milvus cost?
Milvus's paid plans are Standard at From $126 per month, dedicated cluster and Enterprise at From $197 per month, dedicated cluster.
Is Milvus open source?
Yes. Milvus's code is public under the Apache 2.0 licence, so you can run it on your own server.
What are the best Milvus alternatives?
The closest alternatives on Recaply are Pinecone, Qdrant, Supavec, Deeplake and HelixDB.
Where can I try Milvus?
On its official website, milvus.io.
Is Milvus the same as Zilliz Cloud?
No. Milvus is the open source database you install and run yourself. Zilliz Cloud is a fully managed Milvus run by Zilliz, the project's main contributor, with serverless and dedicated clusters on AWS and Google Cloud, and dedicated clusters on Azure.
What hardware does Milvus need?
A Milvus Standalone server needs at least 8 GB of RAM (16 GB suggested), 4 or more CPU cores with SIMD support such as AVX2, an SSD and Docker. A cluster needs at least 32 GB of RAM, and Milvus Lite runs on a laptop.







